高光谱遥感技术能大面积、快速地对岩石类型进行细分，对地质找矿具有重要的指导意义。而核心矿区的控矿岩石由于化学成分及结构接近（类质同象），导致光谱极其相似，给高光谱数据分析带来困难。本项目拟开展面向岩性精细识别的高光谱图像混合像元分解方法研究：首先，设计一种可变端元可变波段的解混思路，将多维光谱立体角波段选择方法嵌入迭代光谱混合分析方法中，实现对每一个像元自动地选择最优端元束和最具区分度波段；然后，深入研究解混过程中的共线性问题，分析波段选择对减轻端元变异性和增大目标端元的区分度的作用，并通过现有数据对解混模型进行精度评价，厘定模型参数，以获得更准确的丰度图。最后，基于高分五号卫星高光谱影像，以祁连山铜铅锌矿区的超基性岩和花岗岩为识别目标，探索上述解混方法区分相似岩性的能力，同时得到岩性的空间分布图，以辅助矿床成因分析，为推动高光谱影像在矿区的应用发展提供理论基础和技术支撑。

The hyperspectral remote sensing can be used for large-scale and fast mapping of rock types, which is important to guide the mineral exploration. However, rocks with mineralization in the core of the ore deposit are very spectrally similar due to their almost identical chemical composition and structure. This project will be focusing on the spectral mixture analysis of hyperspectral image and its application in rock mapping. First, the approach of spectral unmixing using variable endmembers and variable bands is proposed. The band selection method of n-dimensional spectral solid angle (NSSA) is implanted into the process of iterative spectral mixture analysis (ISMA) method, on the purpose of selecting optimal endmember sets and the most discriminating bands. Second, the colinearity in spectral unmixing is lucubrated. The aim is to explore the impact of band selection on endmember variability and target separability. The NSSA-ISMA approach will be evaluated using well-validated datasets. The selection of parameters in NSSA-ISMA will also be discussed in order to achieve more accurate abundance maps. Lastly, this research explores the feasibility of NSSA-ISMA approach in the mapping of ultramafic and granite rocks in the Cu-Pb-Zn mineral deposit in Qilianshan area, based on Gaofen-5 satellite hyperspectral image. As an important result of the spectral unmixing, the abundance maps of key rock units will be used to assist the ore genesis. These three individual researches will provide theoretical and methodological foundations to the widespread use of hyperspectral imaging in mineral exploration.